blob: ae7ff0c24383abfd13ff8bdea937c2aed5049402 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
|
//
// Copyright © 2022-2023 Arm Ltd and Contributors. All rights reserved.
// SPDX-License-Identifier: MIT
//
#include "GpuFsaBackend.hpp"
#include "GpuFsaBackendContext.hpp"
#include "GpuFsaBackendDefaultAllocator.hpp"
#include "GpuFsaBackendId.hpp"
#include "GpuFsaLayerSupport.hpp"
#include "GpuFsaTensorHandleFactory.hpp"
#include "GpuFsaWorkloadFactory.hpp"
#include <armnn/backends/IBackendContext.hpp>
#include <armnn/backends/IMemoryManager.hpp>
#include <Optimizer.hpp>
#include <aclCommon/BaseMemoryManager.hpp>
#include <arm_compute/runtime/CL/CLBufferAllocator.h>
namespace armnn
{
const BackendId& GpuFsaBackend::GetIdStatic()
{
static const BackendId s_Id{GpuFsaBackendId()};
return s_Id;
}
IBackendInternal::IMemoryManagerUniquePtr GpuFsaBackend::CreateMemoryManager() const
{
if (m_UsingCustomAllocator)
{
return std::make_unique<GpuFsaMemoryManager>(m_CustomAllocator);
}
return std::make_unique<GpuFsaMemoryManager>(std::make_unique<arm_compute::CLBufferAllocator>());
}
IBackendInternal::IWorkloadFactoryPtr GpuFsaBackend::CreateWorkloadFactory(
const IBackendInternal::IMemoryManagerSharedPtr& memoryManager) const
{
return std::make_unique<GpuFsaWorkloadFactory>(PolymorphicPointerDowncast<GpuFsaMemoryManager>(memoryManager));
}
IBackendInternal::IWorkloadFactoryPtr GpuFsaBackend::CreateWorkloadFactory(
TensorHandleFactoryRegistry& registry) const
{
std::shared_ptr<GpuFsaMemoryManager> memoryManager;
if (m_UsingCustomAllocator)
{
memoryManager = std::make_shared<GpuFsaMemoryManager>(m_CustomAllocator);
}
else
{
memoryManager = std::make_shared<GpuFsaMemoryManager>(std::make_unique<arm_compute::CLBufferAllocator>());
}
std::unique_ptr<ITensorHandleFactory> factory = std::make_unique<GpuFsaTensorHandleFactory>(memoryManager);
registry.RegisterMemoryManager(memoryManager);
registry.RegisterFactory(std::move(factory));
return std::make_unique<GpuFsaWorkloadFactory>(PolymorphicPointerDowncast<GpuFsaMemoryManager>(memoryManager));
}
IBackendInternal::IWorkloadFactoryPtr GpuFsaBackend::CreateWorkloadFactory(
TensorHandleFactoryRegistry& registry,
const ModelOptions& modelOptions,
MemorySourceFlags inputFlags,
MemorySourceFlags outputFlags) const
{
IgnoreUnused(modelOptions);
// To allow force import if inputFlags/outputFlags are Undefined, set it as Malloc
if (inputFlags == static_cast<MemorySourceFlags>(MemorySource::Undefined))
{
inputFlags = static_cast<MemorySourceFlags>(MemorySource::Malloc);
}
if (outputFlags == static_cast<MemorySourceFlags>(MemorySource::Undefined))
{
outputFlags = static_cast<MemorySourceFlags>(MemorySource::Malloc);
}
std::shared_ptr<GpuFsaMemoryManager> memoryManager;
if (m_UsingCustomAllocator)
{
memoryManager = std::make_shared<GpuFsaMemoryManager>(m_CustomAllocator);
}
else
{
memoryManager = std::make_shared<GpuFsaMemoryManager>(std::make_unique<arm_compute::CLBufferAllocator>());
}
std::unique_ptr<ITensorHandleFactory> factory = std::make_unique<GpuFsaTensorHandleFactory>(memoryManager);
registry.RegisterMemoryManager(memoryManager);
registry.RegisterFactory(std::move(factory));
return std::make_unique<GpuFsaWorkloadFactory>(PolymorphicPointerDowncast<GpuFsaMemoryManager>(memoryManager));
}
std::vector<ITensorHandleFactory::FactoryId> GpuFsaBackend::GetHandleFactoryPreferences() const
{
return std::vector<ITensorHandleFactory::FactoryId> { GpuFsaTensorHandleFactory::GetIdStatic() };
}
void GpuFsaBackend::RegisterTensorHandleFactories(TensorHandleFactoryRegistry& registry)
{
std::shared_ptr<GpuFsaMemoryManager> memoryManager;
if (m_UsingCustomAllocator)
{
memoryManager = std::make_shared<GpuFsaMemoryManager>(m_CustomAllocator);
}
else
{
memoryManager = std::make_shared<GpuFsaMemoryManager>(std::make_unique<arm_compute::CLBufferAllocator>());
}
std::unique_ptr<ITensorHandleFactory> factory = std::make_unique<GpuFsaTensorHandleFactory>(memoryManager);
registry.RegisterMemoryManager(memoryManager);
registry.RegisterFactory(std::move(factory));
}
void GpuFsaBackend::RegisterTensorHandleFactories(TensorHandleFactoryRegistry& registry,
MemorySourceFlags inputFlags,
MemorySourceFlags outputFlags)
{
// To allow force import if inputFlags/outputFlags are Undefined, set it as Malloc
if (inputFlags == static_cast<MemorySourceFlags>(MemorySource::Undefined))
{
inputFlags = static_cast<MemorySourceFlags>(MemorySource::Malloc);
}
if (outputFlags == static_cast<MemorySourceFlags>(MemorySource::Undefined))
{
outputFlags = static_cast<MemorySourceFlags>(MemorySource::Malloc);
}
std::shared_ptr<GpuFsaMemoryManager> memoryManager;
if (m_UsingCustomAllocator)
{
memoryManager = std::make_shared<GpuFsaMemoryManager>(m_CustomAllocator);
}
else
{
memoryManager = std::make_shared<GpuFsaMemoryManager>(std::make_unique<arm_compute::CLBufferAllocator>());
}
std::unique_ptr<ITensorHandleFactory> factory = std::make_unique<GpuFsaTensorHandleFactory>(memoryManager);
registry.RegisterMemoryManager(memoryManager);
registry.RegisterFactory(std::move(factory));
}
IBackendInternal::IBackendContextPtr GpuFsaBackend::CreateBackendContext(const IRuntime::CreationOptions& options) const
{
return IBackendContextPtr{new GpuFsaBackendContext{options}};
}
IBackendInternal::IBackendProfilingContextPtr GpuFsaBackend::CreateBackendProfilingContext(
const IRuntime::CreationOptions&, IBackendProfilingPtr&)
{
return IBackendProfilingContextPtr{};
}
IBackendInternal::ILayerSupportSharedPtr GpuFsaBackend::GetLayerSupport() const
{
static ILayerSupportSharedPtr layerSupport{new GpuFsaLayerSupport};
return layerSupport;
}
std::unique_ptr<ICustomAllocator> GpuFsaBackend::GetDefaultAllocator() const
{
return std::make_unique<GpuFsaBackendDefaultAllocator>();
}
OptimizationViews GpuFsaBackend::OptimizeSubgraphView(const SubgraphView& subgraph,
const ModelOptions& modelOptions) const
{
OptimizationViews optimizationViews(modelOptions);
optimizationViews.AddUntouchedSubgraph(SubgraphView(subgraph));
return optimizationViews;
}
} // namespace armnn
|
